Full text of DOI:10.2165/00019053-200119070-00006

Pharmacoeconomics 2001; 19 (7): 779-790
1170-7690/01/0007-0779/$22.00/0
ORIGINAL RESEARCH ARTICLE
© Adis International Limited. All rights reserved.
Iatrogenic Cost Factors Incorporating
Mild and Moderate Adverse Events in
the Economic Comparison of
Aceclofenac and Other NSAIDs
Francesc Peris,1 Eduard Martínez,1 Xavier Badia2 and Max Brosa3
1
2
Department of Biometry, Medical Division, Almirall Prodesfarma S.A., Barcelona, Spain
Servei d’Epidemiologia Clinica i Salut Publica, Hospital de la Santa Creu i Sant Pau,
Barcelona, Spain
3
Health Outcomes Research Europe, Barcelona, Spain
Objective: To perform a modelled economic analysis of the efficacy and
tolerability of aceclofenac in comparison with those of other nonsteroidal anti-
inflammatory drugs (NSAIDs) used in the treatment of common arthritic disor-
ders including osteoarthritis, rheumatoid arthritis and ankylosing spondylitis.
Design: Adecision analytical model was constructed to represent the clinical and
economic consequences of NSAID treatment. Probabilities of noncompliance,
lack of efficacy and incidence of adverse events were obtained from comparative
randomised double-blind clinical trials. Local unit treatment costs were used and
an expert panel was convened to estimate resource use. Both classical foldback
analysis and bootstrap methods were used to compute point estimates and 95%
confidence limits of costs for NSAID treatment.
Abstract
Patients and interventions: Data were obtained from 12 randomised double-
blind clinical trials included in an earlier meta-analysis.
Main outcome measures: Total costs to the healthcare provider, including
NSAID treatment costs (drug acquisition costs and physician visits for prescription)
and iatrogenic costs (substitution treatment costs for patients not achieving clinical
efficacy and costs of medical visits, treatment, diagnostic tests and hospital stays
associated with adverse events) and the iatrogenic cost factor (ICF) were used as
the primary outcome measures.
Results: Means and 95% confidence intervals revealed no statistically significant
differences in total costs between aceclofenac and other NSAIDs, with the ex-
ception of piroxicam, despite substantial differences in drug acquisition costs.
The ICF for aceclofenac was lower than that for all other comparators, and dif-
ferences in ICF between aceclofenac 200 mg/day and diclofenac 150 mg/day,
indomethacin 100 mg/day, naproxen 1000 mg/day, tenoxicam 20 mg/day or keto-
profen 150 mg/day were statistically significant.
Conclusion: These results show that the comparative overall costs of NSAIDs
bears little relation to drug acquisition cost, and that the ICF is one of the most
important determinants of overall costs.

780
Peris et al.
Nonsteroidal anti-inflammatory drugs (NSAIDs)
NSAIDs for the treatment of osteoarthritis, rheu-
matoid arthritis and ankylosing spondylitis.[8-15]
are widely prescribed for the relief of pain, inflam-
mation and stiffness associated with common ar-
thritic disorders such as osteoarthritis, rheumatoid
arthritis and ankylosing spondylitis. Although sim-
ilar in terms of efficacy, the NSAIDS are notably
different in terms of patient compliance and the
incidence of adverse events.[1] The adverse events
associated with NSAIDs are well known and have
been widely reported in the medical literature.[2,3]
Recent work carried out by McIntosh[4] concluded
that the cost of managing the adverse events asso-
ciated with NSAID use in rheumatoid arthritis was
greater than the cost of drug therapy itself. There-
fore, these costs are likely to be of significant im-
portance when estimating the total costs associated
with any given NSAID therapy. In fact, given the
similarity in efficacy of NSAIDs, the results from
economic analyses are likely to be largely a func-
tion of drug acquisition costs and the comparative
incidence ofadverseeventsassociatedwiththedrugs.
At the same time, studies of the economic im-
pact of NSAID treatment have focused almost ex-
clusively on drug-induced gastroduodenal ulcers,
since these can be directly associated with the treat-
ment and are one of the most severe clinical prob-
lems associated with NSAIDs.[5] However, other
adverse drug reactions such as abdominal pain or
nausea, which are not so severe, are nevertheless
much more frequent,[6] have direct implications
for patients’health-related quality of life[7] and are
likely to incur considerable costs for patients and/or
the healthcare system. A full economic analysis of
NSAIDs should therefore include both the costs of
treating gastroduodenal ulcers and the costs related
to less severe but more frequent adverse events.
The aim of this study was to perform a mod-
elled economic comparison of aceclofenac and
the most frequently used NSAIDs in Europe (diclo-
fenac, naproxen, piroxicam, ketoprofen, tenoxicam
and indomethacin) for the treatment of common
arthritic disorders(osteoarthritis, rheumatoid arthritis
and ankylosing spondylitis). Aceclofenac, 2-[(2, 6-
dichlorophenyl) amino]-phenyl-acetoxyacetic acid, is
an NSAID that is at least as effective as other
Patients and Methods
Model
A decision tree model was developed to reflect
the events related to NSAID use over a 3-month
period of treatment. This time horizon was chosen
because primary data sources were randomised
3- to 6-month clinical trials carried out in the
development of aceclofenac. Figure 1 shows a sin-
gle branch of the model, and represents all conse-
quences after the initiation of treatment with an
NSAID. Cost data are attached to each final state
(outcome) of the model, depending on the pathway
followed by patients reaching that state. To calcu-
late the average (expected) cost for each treatment
option, specific sets of probabilities were obtained
from a series of 12 randomised double-blind clini-
cal trials[1] (see table I).
After a treatment period of 3 months (following
figure 1 from top to bottom), patients can be de-
scribed as: noncompliant, patients who discontinued
therapy because of lack of efficacy, patients who
completed the full treatment period and did not ex-
perience adverse events, or patients who completed
the full treatment period but experienced 1 or more
adverse events. Table II provides a brief descrip-
tion of each branch depicted in figure 1.
As a consequence of previous experience in
clinical trials, the treatment was judged as effective
or not within 21 days of starting treatment, and this
time window was applied to all NSAID options for
the purposes of calculating the cost of replacements.
The occurrence rates of each adverse event obtained
from the 12 randomised clinical trials were used to
compute the probabilities of experiencing the event.
Independence between the probabilities of having 2
different adverse events was assumed; i.e. the prob-
ability of having a second adverse event was assumed
to be independent of the probability of having the
first adverse event. Adverse events were defined as
severe when they led to treatment discontinuation,
irrespective of their relationship with the NSAID.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (7)

Economics of Aceclofenac vs Other NSAIDs
781
NC
LE (DT)
NAE
Patient
Severe (DT)
AP
No second AE
(To E)
MM (CT)
*
C
Severe (DT)
CO
DI
No second AE
(To E)
MM (CT)
*
Severe (DT)
E
*
No second AE
(To E)
MM (CT)
*
Severe (DT)
GU
VO
OR
DE
DZ
HA
No second AE
(To E)
MM (CT)
*
Severe (DT)
AE
No second AE
(To E)
MM (CT)
Severe (DT)
MM (CT)
*
No second AE
(To E)
*
Severe (DT)
No second AE
(To E)
MM (CT)
*
Severe (DT)
No second AE
(To E)
MM (CT)
*
Severe (DT)
MM (CT)
No second AE
(To E)
*
Fig. 1. Decision tree used in the model to estimate clinical and economic consequences of nonsteroidal anti-inflammatory treatment.
AE = adverse events; AP = abdominal pain; C = compliance; CO = constipation; CT = continue treatment; DE = dermatitis; DI =
diarrhoea; DT = discontinue treatment; DZ = dizziness; E = efficacy; GU = gastroduodenal ulcer; HA = headache; LE = lack of
efficacy; MM = mild/moderate; NAE = no adverse events; NC = noncompliance; OR = oesophageal reflux; VO = vomiting; * Apatient
can experience 1 to 4 mild/moderate adverse events.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (7)

782
Peris et al.
Adverse events were classified according to the
double-blind clinical trials included in an earlier
meta-analysis.[1] Table I shows the most relevant
characteristics of these studies. Aggregated data on
noncompliance, lack of efficacy and adverse event
occurrence rates for all studies is shown in table III.
Detailed data for each study on specific adverse
event occurrence, noncompliance and lack of effi-
cacy is available from the authors, on request.
following 9 categories: 6 gastrointestinal catego-
ries (abdominal pain, constipation, diarrhoea, gas-
troduodenal ulcer, vomiting and gastroesophageal
reflux) and 3 nongastrointestinal categories (der-
matitis, dizziness and headache). The grouping of
adverse events into these categories was in accord-
ance with the World Health Organization Adverse
Reaction Dictionary.[16]
Economic Data
The average time to occurrence of the first ad-
verse event was derived from study B1.1.4 (table I).
Thus, the occurrence of the first adverse event as-
sociated with a given NSAID was calculated to be
at day 9 after treatment initiation; second, third and
fourth adverse events were assumed to occur at days
30, 50 and 70, respectively. In order to obtain a better
estimate of the real resources used for each adverse
event, those registered on 2 occasions (even if they
were 2 manifestations of a single continuing symp-
tom episode) were included as 2 separate events.
Resource use associated with each adverse event
was estimated by means of a specially designed sur-
vey among a panel of 10 rheumatologists who had
participated in clinical trials with different NSAIDs.
All physicians completed a questionnaire which
included 2 types of question: the first set of ques-
tions dealt with issues relating to adverse event
treatment, including: specific treatment for the ad-
verse event (drug, dose and treatment duration);
number of medical visits associated with the ad-
verse event (either by phone, at the patient’s home,
ambulatory or in hospital); length of hospital stay
(when indicated); diagnostic tests; the need to dis-
continue treatment with NSAIDs; and, when appli-
cable, the substitution treatment prescribed. The
second set of questions addressed issues such as the
Data Sources
Randomised Clinical Trials
All probabilities used to feed the model (non-
compliance, lack of efficacy and adverse event rates)
were estimated from the results of 12 randomised
Table I. Characteristics of 12 randomised double-blind clinical trials of aceclofenac 100mg twice daily in patients with osteoarthritis, rheumatoid
arthritis or ankylosing spondylitis[1] a
Study number
Indication
Duration
No. of patients entered
(aceclofenac/other NSAID)
Comparative drug
B1.1.4
B1.1.9
B1.1.11
B1.2.4
Q2.4.1
B1.2.5
B1.2.6
B1.2.7
B1.3.1
B1.3.2
B1.3.3
B1.3.4
Osteoarthritis
Osteoarthritis
Osteoarthritis
3 months
3 months
3 months
6 months
3 months
3 months
3 months
3 months
3 months
3 months
3 months
3 months
200/197
190/184
103/102
162/165
153/154
109/110
145/147
87/82
119/123
155/153
60/66
135/138
1618/1621
Diclofenac 50mg tid
Naproxen 500mg bid
Piroxicam 20mg od
Diclofenac 50mg tid
Diclofenac 50mg tid
Indomethacin 50mg bid
Tenoxicam 20mg od
Ketoprofen 50mg tid
Indomethacin 50mg bid
Indomethacin 50mg bid
Naproxen 500mg bid
Tenoxicam 20mg od
Rheumatoid arthritis
Rheumatoid arthritis
Rheumatoid arthritis
Rheumatoid arthritis
Rheumatoid arthritis
Ankylosing spondylitis
Ankylosing spondylitis
Ankylosing spondylitis
Ankylosing spondylitis
Total
a
A further 6-month study (B1.1.6) of aceclofenac versus diclofenac 50mg tid (n = 170/165) was excluded from the economic evaluation
because of the lack of a detailed adverse event database.
bid = twice daily; NSAID = nonsteroidal anti-inflammatory drug; od = once daily; tid = 3 times daily.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (7)

Economics of Aceclofenac vs Other NSAIDs
783
Table II. Definitions of the model pathways
ID
Model states
Description
NC
Noncompliance
Patients who did not complete the follow-up period for any of the following reasons: protocol
violation, default follow-up, patient refusal, did not maintain treatment, and other (patients who
withdrew because of adverse events or lack of efficacy were not included here)
C
LE
E
Compliance
Lack of efficacy
Efficacy
All patients except noncompliant patients
Patients who withdrew from treatment because of lack of efficacy (deterioration)
Compliant patients who showed a positive response to treatment (patients with and without
adverse events were included)
NAE
AE
No adverse event
Adverse event
Patients who completed the follow-up period without adverse events
Patients who experienced at least 1 adverse event (irrespective of whether or not they
discontinued the treatment). The model allowed for patients to experience from 1 to 4 different
mild/moderate adverse events. Patients who experienced a severe adverse event discontinued
the treatment
AP
CO
DI
GU
VO
OR
DE
DZ
HA
Abdominal pain
Constipation
Diarrhoea
Gastroduodenal ulcer
Vomiting
Oesophageal reflux
Dermatitis
Dizziness
Patients who experienced an abdominal pain episode
Patients who experienced a constipation episode
Patients who experienced a diarrhoea episode
Patients who developed a gastroduodenal ulcer
Patients who experienced a vomiting episode
Patients who experienced an oesophageal reflux episode
Patients who experienced a dermatitis episode
Patients who experienced an episode of dizziness
Patients who experienced a headache episode
Headache
time needed to diagnose lack of efficacy and sub-
stitute treatments. Median values were used to
describe the medical resources associated with
treatment of adverse events and table IVshows the
daily cost of mild/moderate and severe adverseevents
estimated from the clinicians’responses.
NSAID daily acquisition costs (table V) were
calculated as an average value of the market price
of each brand of a given NSAID available in Spain,
weighted by the market share of each brand name.[17]
The cost associated with lack of efficacy (substitu-
tion treatment cost) was estimated to be $US0.42/day
for all NSAID options. Nondrug costs (diagnostic
tests, medical visits and hospitalisation) were ob-
tained from hospital accounting data sets[18] when
available, and from other sources.[19,20] All costs
are in $US, 1996 values [$US1 = 166.08 Spanish
pesetas (Pta)]. In all analyses, costs and not charges
were used. The study was conducted from the per-
spective of the healthcare system.
NSAID cost + iatrogenic costs
ICF =
NSAID cost
‘NSAID cost’ represents costs which are strictly
related to NSAID acquisition and administration,
that is, NSAID daily cost by duration of therapy
and the cost of medical visits (prescription visit).
‘Iatrogenic costs’ are the costs related to treatment
of adverse events and substitution treatment after
NSAID discontinuation for any cause. Alower ICF
means a ‘better’ option, i.e. an option with lower
additional costs for treating adverse events. For
example, an ICF of 1.5 means that for each dollar
spent on the drug it will be necessary to spend an
additional 0.5 dollars on treating adverse events.[21,22]
Statistical Analysis
In order to test the robustness of the results, 2
different statistical approaches were used to esti-
mate the expected costs and ICF associated with
each NSAID option: 1) foldback analysis and 2) a
parametric bootstrap method.[23,24]
Economic Analysis
Iatrogenic Cost Factor Analysis
The following formula was used to estimate the
iatrogenic cost factor (ICF):
In both methods, the expected costs (NSAID
cost, iatrogenic costs and total costs) and ICFs were
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (7)

784
Peris et al.
Table III. Aggregated results from 12 randomised double-blind clinical trials of aceclofenac in patients with osteoarthritis, rheumatoid arthritis
or ankylosing spondylitis[1]
Median occurrence rates (range)
AC (n = 12) DC (n = 3)
IN (n = 3)
9.2%
NX (n = 2)
12.3%
TX (n = 2)
9.5%
KT (n = 1)
7.3%
PX (n = 1)
6.9%
Noncompliance
Compliance
8.4%
13.0%
(2.1-17.9)
91.6%
(82.1-97.9)
6.2%
(0.0-9.2)
85.1%
(78.9-92.6)
63.9%
(26.1-84.5)
24.6%
(7.7-56.3)
10.2%
(5.0-21.9)
0.7%
(0.0-4.6)
1.8%
(0.0-6.5)
0.0%
(0.0-1.8)
2.5%
(0.0-7.7)
3.6%
(0.5-10.4)
2.1%
(0.0-6.5)
0.8%
(0.0-7.4)
0.0%
(3.0-17.3)
87.0%
(82.7-97.0)
9.1%
(4.1-13.0)
78.7%
(74.0-87.9)
35.7%
(27.4-61.2)
38.3%
(26.7-51.3)
17.9%
(12.8-21.1)
1.9%
(1.0-3.0)
5.1%
(2.8-9.5)
0.5%
(0.0-0.5)
5.6%
(0.5-7.0)
5.5%
(4.1-5.7)
1.4%
(0.5-3.1)
1.4%
(1.0-4.0)
0.0%
(6.4-12.2)
90.8%
(87.8-93.6)
4.6%
(4.1-9.1)
84.5%
(83.7-86.3)
40.5%
(22.8-48.2)
45.8%
(36.3-60.9)
14.0%
(9.3-20.8)
0.5%
(0.0-2.4)
1.8%
(0.0-7.2)
0.0%
(0.0-0.0)
8.0%
(1.8-10.1)
6.4%
(2.3-13.6)
2.3%
(0.0-2.3)
10.8%
(5.9-15.2)
1.6%
(6.1-18.5)
87.7%
(81.5-93.9)
2.9%
(2.7-3.0)
84.8%
(78.8-90.9)
57.4%
(57.1-57.6)
27.5%
(21.7-33.3)
15.8%
(12.8-18.9)
2.7%
(1.1-4.4)
0.6%
(0.0-1.1)
0.0%
(0.0-0.0)
4.4%
(3.3-5.6)
3.1%
(1.7-4.4)
0.6%
(0.0-1.1)
0.3%
(0.0-0.6)
0.0%
(8.8-10.1)
90.6%
(89.9-91.2)
6.6%
(5.1-8.2)
83.9%
(83.0-84.8)
59.7%
(58.5-60.9)
24.2%
(23.9-24.5)
17.3%
(16.6-17.9)
0.0%
(0.0-0.0)
0.8%
(0.7-0.9)
0.6%
(0.4-0.7)
1.8%
(0.9-2.7)
1.0%
(0.7-1.3)
0.2%
(0.0-0.4)
1.1%
(0.0-2.2)
1.6%
92.7%
13.4%
79.3%
48.7%
30.5%
18.5%
0.0%
93.1%
1.0%
92.1%
77.3%
14.7%
9.6%
0.0%
1.3%
0.0%
1.3%
2.6%
0.0%
0.0%
0.0%
Lack of efficacy
Efficacy
No adverse events
Adverse events
Abdominal pain
Constipation
Diarrhoea
2.2%
Gastroduodenal ulcer
Vomiting
2.2%
2.2%
Oesophageal reflux
Dermatitis
2.2%
2.2%
Dizziness
0.0%
Headache
1.1%
(0.0-2.2)
(0.0-0.0)
(0.8-5.9)
(0.0-0.0)
(0.4-2.7)
AC = aceclofenac 200 mg/day; DC = diclofenac 150 mg/day; IN = indomethacin 100-150 mg/day; KT = ketoprofen 150 mg/day; n = number
of trials; NX = naproxen 1000 mg/day; PX = piroxicam 20 mg/day; TX = tenoxicam 20-40 mg/day.
estimated using the same set of probabilities from
the 12 randomised clinical trials mentioned earlier.
The conditional probabilities of mild/moderate
and severe adverse events, given that the adverse
event has occurred, were derived from the study
B1.1.4. The expected number of adverse events per
patient (i.e. the mean number of adverse events
per patient with at least 1 adverse event) were com-
puted for each group of adverse events and rand-
omised clinical trial. This figure was used to
estimate the probability of experiencing successivead-
verseevents,giventhatamild/moderate adverse event
has occurred. This probability was adjusted for pa-
tients who discontinued the treatment because of
severe adverse events.
Foldback Analysis
For each study and NSAID option, the expected
costs (NSAID cost, iatrogenic costs and total costs)
were estimated for each of 14 560 tree outcomes of
the model (fig. 1). The means were estimated by
averaging all expected costs, and variances were
estimated assuming a multinomial distribution for
all tree outcomes. The ICFs for each of the tree
outcomes were obtained using the values of ex-
pected costs described above. To take into account
the skewed distribution of ICF values the data were
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (7)

Economics of Aceclofenac vs Other NSAIDs
785
Table IV. Costs of mild/moderate and severe adverse events (AEs)
associated with nonsteroidalanti-inflammatory drugsinpatientswith
osteoarthritis, rheumatoid arthritis or ankylosing spondylitis as esti-
mated by a panel of 10 rheumatologists
log-transformed and the mean and variance of ICFs
for each study and NSAID option were estimated
using transformed values and parametric methods.
Means and confidence interval (CI) values were
then back-transformed to give estimates for the ICFs.
Median costs in $USa (range)
mild/moderate AEs severe AEs
Abdominal pain
Medical visits
AE treatment
Diagnostic tests
Hospital stay
37.5
323.5
Bootstrap Estimate Procedure
10.9 (0.0-31.8)
20.3 (0.0-126.3)
0.0 (0.0-65.0)
46.3 (14.1-1691.0)
62.4 (23.1-154.4)
68.8 (4.6-137.5)
146.4 (0.0-253.5)
To test the robustness of the results, NSAID cost,
iatrogenic costs, total costs and ICFs were also
estimated using a parametric bootstrap procedure[23]
(see technical appendix). The NSAID cost and iatro-
genic costs per patient were simulated using a
marginal multinomial distribution that adequately
describes the skewness of the adverse events costs as-
sociated with NSAIDs and provides an easier means of
calculating confidence intervals.
Constipation
Medical visits
AE treatment
Diagnostic tests
Hospital stay
3.3
192.6
37.0 (3.3-479.0)
4.6 (0.0-23.0)
17.7 (0.0-89.5)
126.5 (0.0-189.4)
1.6 (0.0-82.7)
0.0 (0.0-9.5)
0.0 (0.0-10.9)
Diarrhoea
Medical visits
AE treatment
Diagnostic tests
Hospital stay
12.5
10.9 (0.0-31.6)
0.0 (0.0-19.0)
0.0 (0.0-35.4)
697.0
673.2 (14.1-1808.0)
4.9 (0.0-34.5)
17.7 (4.6-134.8)
0.0 (0.0-324.0)
Results
Table VI shows the total costs per patient who
received treatment (NSAID cost + iatrogenic costs)
and 95% CI for each randomised clinical trial and
NSAID option. The differences between treatment
options (aceclofenac total costs minus comparator
total costs) and 95% CI are shown. In most studies,
the differences were >0 (indicating higher costs for
aceclofenac), ranging from –16.4 (study Q2.4.1;
aceclofenac vs diclofenac) to 50.7 (study B1.3.1;
aceclofenac vs indomethacin) US dollars per pa-
tient. With respect to the 95% CIs for total costs,
of the 12 individual studies included in the eco-
nomic evaluation, 6 showed 95% CIs that included
0 (i.e. differences were not statistically significant)
and 6 showed 95% confidence limits that did not
include 0.
When results from trials using the same compa-
rator were aggregated it was found that most dif-
ferences between aceclofenac and the compara-
tors were >0, ranging from –13.9 (aceclofenac vs
ketoprofen) to 24.0 (aceclofenac vs piroxicam) US
dollars per patient. However, most of the 95% CIs
included 0, i.e. there were no statistically significant
differences between the total costs of aceclofenac,
diclofenac, indomethacin, naproxen, tenoxicam
and ketoprofen. Total costs associated with aceclo-
fenac use were significantly higher than total costs
for piroxicam.
Gastroduodenal ul- 409.8
cer
2096.0
Medical visits
AE treatment
Diagnostic tests
Hospital stay
210.5 (81.7-3467.0) 1215.1 (50.1-2657.2)
63.1 (29.5-126.3)
72.8 (4.6-145.6)
72.8 (46.4-126.3)
136.0 (72.8-157.1)
765.7 (647.9-1133.8)
Vomiting
13.4
1063.6
Medical visits
AE treatment
Diagnostic tests
13.4 (0.0-50.1)
1.2 (0.0-4.3)
0.0 (0.0-65.0)
988.7 (21.7-1807.9)
4.2 (0.8-668.2)
68.7 (0.0-102.5)
Hospital stay
0.0
Oesophageal reflux 30.1
335.2
Medical visits
AE treatment
Diagnostic tests
Hospital stay
7.1 (0.0-42.5)
10.6 (0.0-378.8)
0.0 (0.0-65.0)
39.2 (21.7-578.6)
94.7 (38.6-378.8)
67.3 (65.0-137.8)
134.0 (0.0-179.3)
Dermatitis
15.3
162.7
Medical visits
AE treatment
Diagnostic tests
Hospital stay
14.1 (0.0-82.3)
1.3 (0.0-16.4)
0.0 (0.0-9.2)
88.8 (21.7-342.0)
2.7 (0.0-18.9)
7.8 (0.0-14.1)
57.9 (0.0-133.8)
Dizziness
17.1
134.4
Medical visits
AE treatment
Diagnostic tests
Hospital stay
10.9 (0.0-56.7)
1.7 (0.0-16.6)
0.0 (0.0-9.9)
83.3 (10.9-2435.2)
4.1 (0.0-18.4)
4.8 (0.0-132.1)
34.5 (0.0-121.1)
Headache
4.0
329.5
Medical visits
AE treatment
Diagnostic tests
Hospital stay
3.3 (0.0-31.6)
0.6 (0.0-2.9)
0.0 (0.0-98.3)
76.2 (28.3-354.7)
5.4 (0.0-11.0)
100.5 (4.6-112.1)
156.2 (0.0-303.1)
a
$US1 = 166.08 Spanish pesetas (1996 values).
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (7)

786
Peris et al.
Table V. Daily acquisition costs ($USa) of aceclofenac and other
common nonsteroidal anti-inflammatory drugs (NSAIDs) used in 12
randomised double-blind trials of patients with osteoarthritis, rheu-
matoid arthritis or ankylosing spondylitisb
Studies which have examined iatrogenic cost fac-
torsfordifferent NSAIDshavemostlyconcentrated on
costs associated with more severe adverse events,
particularly gastric ulcers[5,21,26] and have generally
ignored other milder, but more frequent adverse
events. One of the few studies to include milder
events found that taking them into account could
have considerable effect on the relative performance
of certain NSAIDs.[27] This is one of the first stud-
ies to take into account the considerable costs to
the healthcare system associated with milder but
more frequent adverse events such as abdominal
pain, dermatitis, headache, etc. In our study, be-
tween 11 and 22% of the costs of adverse events
associated with NSAIDs were accounted for by these
milder adverse events.
The results from this economic evaluation re-
veal that, although the daily cost of aceclofenac is
higher than that of any of the other comparator
treatments, the overall cost of treatment is very sim-
ilar across the whole range of NSAIDs, primarily be-
cause the ICF associated with aceclofenac was
significantly lower than the ICFs of the other NSAID
options considered with the exception of piroxicam.
The differences in ICF values are likely to be caused
by the differences in the cost of treating mild and
moderate adverse events associated with the use of
different NSAIDs, as aceclofenac use was gener-
ally associated with a lower frequency of most of
these adverse events (table III).
It should be pointed out that despite the better
ICF for aceclofenac compared with the other op-
tions tested here, the total cost of treatment (NSAID
cost + iatrogenic cost) still tended to be slightly higher
for aceclofenac, particularly in comparison with in-
domethacin and naproxen, because of the higher ac-
quisition costs of aceclofenac. Nevertheless, these
differences were not statistically significant with
the exception of piroxicam.
It should also be borne in mind that the analysis
performed here did not take into account either
indirect costs sustained by patients, or intangible
costs associated with treatment, such as the pain
and discomfort borne by patients. The quality of
life of patients who received NSAIDs is sensitive
NSAID (mg/day)
Aceclofenac 200
Diclofenac 150
Indomethacin 100
Ketoprofen 150
Naproxen 1000
Piroxicam 20
Daily cost
0.73
0.49
0.18
0.26
0.44
0.40
Tenoxicam 20
0.43
a
b
$US1 = 166.08 Spanish pesetas (1996 values).
See table I.
Table VII shows the ICF values and 95% CIs for
each randomised clinical trial and NSAID option.
The ICF ratios between treatment options (ICF of
:
aceclofenac ICF of comparator) and 95% CIs are
also shown. In most individual studies, the ICF ra-
tios and 95% CIs were <1. The majority of pooled
ICF ratios and 95% CIs were also <1. These results
indicate that the ICF of aceclofenac was signifi-
cantly lower than the ICF of diclofenac, indometh-
acin, naproxen, tenoxicam and ketoprofen. The ICF
of aceclofenac was similar to that of piroxicam.
The expected costs (iatrogenic costs and total
costs) estimated by means of 2 approaches (foldback
and bootstrap methods) were comparable; the ICF
ratios were similar in both statistical approaches
(table VIII). Although the ranges of the 95% CIs
estimated by means of foldback analysis were nar-
rower than those estimated by means of the boot-
strap procedure, a similar trend can be observed
with both methods.
Discussion
Any economic analysis of the use of NSAIDs
should take into account the considerable adverse
effects associated with these drugs. The fact that
the costs of treating adverse events associated with
NSAIDs may be greater than the cost of the treat-
ment itself[4] makes the study of iatrogenic cost
factors extremely important.[5] This is especially true
when most NSAIDs analysed in this study have been
shown to be largely similar in terms of efficacy.[1,25]
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (7)

Economics of Aceclofenac vs Other NSAIDs
787
to adverse events and its assessment could increase
the differences between the drugs when safety pro-
files are not similar.[7] Taking these factors into ac-
count might well favour aceclofenac, because of its
better adverse event profile. In the present study,
these costs were not included as there were logistic
restrictions on the amount of data which could be
collected. Broadening the focus of the analysis in
this way could provide an avenue for future stud-
ies.
Although the data used in this study were ob-
tained from randomised clinical trials, further stud-
ies are required to determine efficacy in real-world
clinical practice conditions, where issues such as
treatment compliance become considerably more
important. Thus, although little difference was found
between NSAIDs in terms of short term efficacy
under the ideal conditions of randomised clin-
ical trials, differences may be found in longer term
studies of effectiveness; further research is needed
in this area. Furthermore, the restriction of the treat-
ment period to 3 months in the analysis (because of
follow-up periods used in the randomised clinical
trials) ignores adverse events that may appear after
the 3-month period. Given that patient randomisa-
tion compensates for the possible bias of patient
idiosyncrasy regarding susceptibility to NSAIDs,
the short study time-horizon may be advantageous
to those NSAID options with a worse iatrogenic
profile, since these NSAIDs could also show the
highest probabilities of adverse event and related
costs, given a longer time horizon.[26]
Table VI. Total costs ($USa) per patient receiving treatment (NSAID cost + iatrogenic costs) for aceclofenac 200 mg/day and other common
NSAIDs used in 12 randomised double-blind trials in patients with osteoarthritis, rheumatoid arthritis or ankylosing spondylitisb
Study referenceb
Costs for aceclofenac (95% CI)
Costs for comparator drug (95% CI) Differencec (95% CI)
Diclofenac 150 mg/day
B1.1.4
B1.2.4
Q2.4.1
123.8 (121.3,126.3)
97.0 (95.1,98.9)
104.9 (102.0,108.3)
108.6 (95.8,121.3)
123.6 (120.4,126.9)
74.7 (73.2,76.3)
121.3 (117.3,125.3)
106.5 (98.1,115.1)
0.2 (–18.2,18.6)
22.2 (13.2,31.3)
–16.4 (–39.6,6.8)
2.0 (–13.0,17.0)
Mean value
Indomethacin 100 mg/day
B1.2.5
B1.3.1
B1.3.2
78.4 (76.6,80.2)
86.8 (82.4,91.1)
99.9 (96.2,103.6)
77.0 (74.7,79.3)
87.9 (84.4,91.4)
–8.3 (–36.9,20.2)
50.7 (19.2,82.3)
25.5 (18.8,32.1)
22.6 (–23.3,68.6)
150.7 (145.1,156.2)
102.5 (100.4,104.6)
110.5 (76.6,144.5)
Mean value
Naproxen 1000 mg/day
B1.1.9
B1.3.3
86.8 (84.9,88.6)
113.8 (107.1,120.5)
100.3 (81.5,119.0)
63.9 (62.0,65.8)
92.5 (86.5,98.6)
78.2 (68.3,88.1)
22.9 (20.6,25.1)
21.3 (10.4,32.1)
22.1 (–15.8,60.0)
Mean value
Tenoxicam 20 mg/day
B1.2.6
B1.3.4
Mean value
90.5 (88.9,92.2)
85.4 (83.9,86.8)
87.9 (84.4,91.5)
90.1 (87.2,93.0)
83.3 (80.1,86.5)
86.7 (84.4,89.1)
0.4 (–20.5,21.3)
2.1 (–21.5,25.6)
1.2 (–0.5,3.0)
Ketoprofen 150 mg/day
B1.2.7
94.8 (91.1,98.5)
108.7 (100.4,117.0)
63.4 (60.9,66.0)
–13.9 (–61.1,33.3)
24.0 (20.0,28.0)
Piroxicam 20 mg/day
B1.1.11
87.4 (84.9,89.9)
a
b
c
$US1 = 166.08 Spanish pesetas (1996 values).
See table I.
Total costs of aceclofenac minus total costs of comparator drug.
CI = confidence intervals; NSAID = nonsteroidal anti-inflammatory drug.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (7)

788
Peris et al.
The fact that ICFs were similar using both clas-
Despite the use of both the foldback analysis and
bootstrap methods, a potential limitation of this study
is that no classic sensitivity analysis was performed.
Without such an analysis it is difficult to determine
the extent to which results are dependent on factors
which are susceptible to inherent variation or lack of
precision. However, 95% CIs were obtained in or-
der to assess variability of input variables and these
have shown the robustness of the results from the
analysis.
A further weakness would be the reliance on
expert panels for the generation of data on resource
use. The risks associated with the use of such pan-
els has been well documented,[28] though every ef-
fort was made to avoid these problems by incorpo-
rating experts from different healthcare sectors and
by using structured elicitation procedures.
sic foldback analysis and bootstrap methods sug-
gests that the results are robust, especially as the
analytical approach used in the 2 methods is some-
what different. While the foldback technique cal-
culates the expected value of a certain variable from
probabilities of occurrence of different events (in this
case the cost associated with different treatment
outcomes), the bootstrap method generates a sam-
pling distribution from probabilities (and associ-
ated costs) derived from meta-analyses,[24] and there-
fore takes into account the variability and skewness
of input probabilities. Nevertheless, if the results are
to be extrapolated to other conditions, factors such
as compliance rates, which differed slightly among
the 3 conditions in which the NSAIDs were tested,
should also be taken into account.
Table VII. Iatrogenic cost factor (ICF) and 95% confidence intervals (CI) for patients receiving aceclofenac 200 mg/day or other common
nonsteroidal anti-inflammatory drugs in 12 randomised double-blind trials in patients with osteoarthritis, rheumatoid arthritis or ankylosing
spondylitisa
Study referencea
Aceclofenac ICF (95% CI)
Comparator drug ICF (95% CI)
Ratio (95% CI)b
Diclofenac 150 mg/day
B.1.1.4
B.1.2.4
Q.2.4.1
Mean value
1.68 (1.53,1.83)
1.46 (1.30,1.61)
1.46 (1.29,1.63)
1.53 (1.41,1.65)
1.95 (1.79,2.10)
1.57 (1.42,1.73)
2.13 (1.96,2.30)
1.88 (1.80,1.97)
0.86 (0.75,0.97)
0.93 (0.82,1.03)
0.69 (0.58,0.79)
0.82 (0.71,0.94)
Indomethacin 100-150 mg/day
B.1.2.5
B.1.3.1
B.1.3.2
1.21 (1.02,1.40)
1.83 (1.65,2.01)
1.52 (1.35,1.69)
1.52 (1.24,1.80)
2.10 (1.91,2.30)
2.43 (2.25,2.60)
2.16 (1.99,2.32)
2.23 (2.18,2.28)
0.58 (0.46,0.70)
0.75 (0.64,0.87)
0.70 (0.60,0.81)
0.68 (0.59,0.76)
Mean value
Naproxen 1000 mg/day
B.1.1.9
B.1.3.3
1.20 (1.04,1.35)
1.57 (1.31,1.82)
1.38 (1.13,1.64)
1.36 (1.20,1.52)
1.74 (1.49,1.98)
1.55 (1.42,1.68)
0.88 (0.77,0.99)
0.90 (0.72,1.08)
0.89 (0.88,0.91)
Mean value
Tenoxicam 20-40 mg/day
B.1.2.6
B.1.3.4
Mean value
1.42 (1.26,1.59)
1.34 (1.17,1.51)
1.38 (1.32,1.44)
1.75 (1.59,1.92)
1.50 (1.33,1.66)
1.63 (1.54,1.71)
0.81 (0.70,0.92)
0.89 (0.77,1.02)
0.85 (0.80,0.91)
Ketoprofen 150 mg/day
B.1.2.7
1.30 (1.09,1.52)
2.26 (2.04,2.48)
1.32 (1.12,1.51)
0.58 (0.45,0.71)
0.89 (0.75,1.02)
Piroxicam 20 mg/day
B.1.11
1.17 (0.98,1.37)
a
b
See table I.
Aceclofenac ICF : comparator drug ICF.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (7)

Economics of Aceclofenac vs Other NSAIDs
789
Table VIII. Comparison of iatrogenic cost factor and 95% CIs for
aceclofenac 200 mg/day and other common nonsteroidal anti-in-
flammatory drugs used in 12 randomised double-blind trials in
patients with osteoarthritis, rheumatoid arthritis or ankylosing spon-
dylitis;a calculated using 2 statistical approaches
tribution that adequately describes the skewness of
the adverse events costs associated with NSAIDs.
Let Bk denote a simulated sample of patients with
the same sample size as the randomised clinical
trial. The estimates of NSAID cost, iatrogenic costs
and total cost of this bootstrap sample Bk are de-
noted by NCBk, AECBk and TCBk, respectively, and
are obtained by averaging individual NSAID costs,
iatrogenic costs and total costs for all simulated
patients. The ICF for the same simulated sample Bk
is denoted ICFBk and is obtained by averaging the
individual ICF values for all simulated patients in
the sample Bk.
Comparator drug
(mg/day)
Statistical approach
Foldback (95% CI) Bootstrap (95% CI)
Diclofenac 150
Indomethacin 100-150 0.68 (0.59,0.76)
Naproxen 1000
Tenoxicam 20-40
Ketoprofen 150
Piroxicam 20
0.82 (0.71,0.94)
0.82 (0.71,0.95)
0.59 (0.49,0.72)
0.89 (0.71,1.10)
0.86 (0.72,1.02)
0.57 (0.39,0.84)
0.88 (0.69,1.13)
0.89 (0.88,0.91)
0.85 (0.80,0.91)
0.58 (0.45,0.71)
0.89 (0.75,1.02)
a
See table I.
CI = confidence interval.
Let K denote the total number of independent
bootstrap simulated samples (B1,...,BK). The boot-
strap estimate (d*) of parameter d is then defined
as:
The recent appearance of cyclo-oxygenase
(COX)-2 inhibitors,[29,30] which have improved safety
profiles, means that these drugs should be taken
into account in any future analysis of the costs
associated with the pharmaceutical treatment of
osteoarthritis, rheumatoid arthritis and ankylosing
spondylitis. Future studies in this area should also
take into account the use of prophylactic treatment
for gastrointestinal events, which could also affect
results, particularly if treatment was more effective
for some NSAIDs than for others.
k
1
K
∗
d =
d Bj
å
j = 1
where d is NC, AEC, TC or ICF and K = 10 000
bootstrap replications or independent bootstrap
simulated samples and j denotes the jth sample.
The bootstrap variance of d* is defined as:
K
[d Bj − d ∗]2
(K − 1)
Conclusion
Var (d∗) =
å
j = 1
This study has shown that when costs deriving
from a wide range of adverse events associated
with NSAID use are included in the economic
analyses of these drugs, drug acquisition costs
play only a very limited role in determining the
overall costs of treatment. In this particular analy-
sis, aceclofenac proved to have a better ICF than the
other frequently used NSAIDs included in the anal-
ysis (with the exception of piroxicam, versus
which aceclofenac had significantly higher total
costs and a similar ICF), which led to very similar
overall costs, despite a higher initial acquisition cost.
where d is NC, AEC, TC or ICF and K = 10 000
bootstrap replications or independent bootstrap sim-
ulated samples. This variance was used to estimate
the 95% CIs for all parameters.
References
1. Peris F, Bird HA, Serni U, et al. Treatment compliance and
safety of aceclofenac versus standard NSAIDS in patients
with common arthritic disorders: a meta-analysis. Eur J
Rheumatol Inflamm 1996; 16: 37-45
2. Brand KD. Should nonsteroidal anti-inflammatory drugs be used
to treat osteoarthritis? Rheum Dis Clin North Am 1993; 19:
29-44
3. Scholes D, Stergachis A. Nonsteroidal anti-inflammatory
discontinuation in patients with osteoarthritis. J Rheumatol
1995; 22: 708-12
4. McIntosh E. The cost of rheumatoid arthritis. Br J Rheumatol
1996; 35: 781-90
Technical Appendix
Description of the Bootstrap Procedure
5. De Pouvourville G. The economic consequences of NSAID-
induced gastropathy: the French context. Scand J Rheumatol
1992; 96 (Suppl): 49-53
The NSAID cost and iatrogenic costs per patient
were simulated using a marginal multinomial dis-
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (7)

790
Peris et al.
6. Zeidler H. Epidemiology of NSAID-induced gastropathy. Clin
21. De Pouvourville G. The iatrogenic cost of nonsteroidal anti-in-
flammatory drug therapy. Br J Rheumatol 1995; 34 Suppl. 1:
19-24
22. Evans KW, Boan JA, Evans JL, et al. Economic evaluation of
oral sumatriptan compared with oral caffeine/ergotamine for
migraine. Pharmacoeconomics 1997; 12: 565-77
23. Efron B, Tibshirami R. An introduction to the bootstrap. Chap-
man and Hall: New York, 1993
24. Pasta DJ, Taylor JL, Hening JM. Probabilistic sensitivity anal-
ysis incorporating the bootstrap: an example comparing treat-
ments for the eradication of helicobacter pylori. Med Decis
Making 1999; 19: 353-63
25. Wynne HA, Campbell M. Pharmacoeconomics of nonsteroi-
dal anti-inflammatory drugs (NSAIDs). Pharmacoeconomics
1994; 3: 107-23
26. MacDonald TM, Morant SV, Robinson GC, et al. Association
of upper gastrointestinal toxicity of non-steroidal anti-inflam-
matory drugs with continued exposure: cohort study. BMJ
1997; 315: 1333-7
27. McAbe CJ, Akehurst RL, Kirsch J, et al. Choice of NSAID and
management strategy in rheumatoid arthritis and osteoar-
thritis: the impact on costs and outcomes in the UK. Phar-
macoeconomics 1998; 14: 191-9
28. Evans C. The use of consensus methods and expert panels in
pharmacoeconomic studies: practical applications and meth-
odological shortcomings. Pharmacoeconomics 1997; 12: 121-9
29. Simon LS, Weaver AL, Graham DY, et al. Anti-inflammatory
and upper gastrointestinal effects of celecoxib in rheuma-
toid arthritis: a randomized controlled trial. JAMA1999; 282:
1921-8
30. Langman MJ, Jensen DM, Watson DJ. Adverse upper gastroin-
testinal effects of rofecoxib compared with NSAIDs. JAMA
1999; 282: 1929-33
Rheumatol 1991; 10: 369-73
7. Nichol MB, Harada ASM. Measuring the effects of medication
use on health-related quality of life in patients with rheuma-
toid arthritis. Pharmacoeconomics 1999; 16 (5 Pt 1): 433-48
8. Torrejón VA. Treatment of severe rheumatic pain with aceclo-
fenac: results of comparative study versus placebo. Acta The-
rapeutica 1988; 14: 275-80
9. Díaz C, Belmonte MA, Olmedo J, et al. Eficacia y tolerancia de
aceclofenaco en el tratamiento de la gonartrosis. Revista Es-
panola de Reumatologia 1992; 19: 142-6
10. Giorgianni G, Ottaviani C, Soliano A, et al. Efficacy and toler-
ability of aceclofenac versus ketoprofen in the treatment of
rheumatoid arthritis. Curr Ther Res Clin Exp 1992; 51: 175-84
11. Pasero G, Ruju G, Marcolongo R. Aceclofenac versus naproxen
in the treatment of AS: a double-blind, controlled study. Curr
Ther Res Clin Exp 1994; 55: 833-42
12. Birrell DH, Roma J, Bowdler JM. Evaluation of the efficacy and
safety of aceclofenac in the treatment of osteoarthritis. Br J
Clin Res 1995; 6: 45-55
13. Martin-Mola E, Gijón-Baños J, Ansoleaga J. Aceclofenac in
comparison to ketoprofen in the treatment of rheumatoid
arthritis. Rheumatol Int 1995; 15: 111-6
14. Ward DE, Veys EM, Bowdler JM, et al. Comparison of aceclo-
fenac with diclofenac in the treatment of osteoarthritis. Clin
Rheumatol 1995; 14: 656-62
15. Dawson AJ, Roma J, Bowdler JM. Aceclofenac in the
treatment of rheumatoid arthritis: a multicentre, double-blind,
randomised, placebo controlled study. Eur J Rheumatol Inflamm
1996; 16: 23-7
16. WHO Collaboration Centre for International Drug Monitoring.
Adverse reaction dictionary. Uppsala: World Health Organi-
zation (WHO), 1995
17. International Marketing Services. Annual Report. IMS, Sep-
tember 1995
18. De Bellvitge H. Tarifa de pruebas y procedimientos del Hospital
de Bellvitge. Barcelona: Hospital de Bellvitge, 1996
19. Garcia Cardona F. Análisis de la variabilidad de los costes en
atención primaria. Valencia: Resumen de las XV Jornadas de
Economia de la Salud, 1994
20. Análisis comparativo por grupos de centros. Dirección General
de Regimen Económico. Valencia:GeneralitatValenciana, 1995
Correspondence and offprints: Dr. Xavier Badia, Servei
d’Epidemiologia Clinica i Salut Publica, Edifici UAB-Casa
deConvalescencia, SantAntoniMariaClaret171, 08041Bar-
celona, Spain.
E-mail: xbadia@cochrane.es
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (7)